Research map / Detection and segmentation / 3D detection and driving
Depth-based and label-free 3D detection
Detect 3D objects using pseudo-LiDAR depth, stereo, and label-free mobile object discovery.
58 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2019–2026) · an approach in 3D detection and driving, keyword cluster “3d object · monocular · kitti”
State History Benchmarks Tried Checklist Preprints Who Export
Where it stands
Still moving The best reported result is still improving on 1 of 3 tracked benchmarks.
22 papers in 2023–24 and 15 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic shrank, from 13.9% to 9.9%.
Open questions the data points to
- nuScenes val (mAP): best went from 63.6 (up to 2024) to 68.72 in 2025 — ViKIENet: Towards Efficient 3D Object Detection with Virtual Key Instance Enhanced Network (CVPR 2025 · 10 citations)
No longer reported by papers here: NuScenes validation (mAP, last in 2024); nuScenes test (mAP, last in 2024).
How this is decided
A benchmark is tracked when at least four papers of this approach report their own method on the same dataset and metric in at least two different years. Its record is the best number reported up to 2024 compared with the best reported in 2025–26: still improving means at least 1% better, flat means at most 0.2% better with at least two papers from 2025–26 trying. Numbers on a different scale from the rest (a fraction among percentages) and far outliers are left out. Results come from different papers with different setups, so they are signals, not head-to-head comparisons. The share of papers is counted at ICML, NeurIPS, ICLR, CVPR and AAAI.
How it got here
Most cited paper of each year
- 2019 · Pseudo-LiDAR From Visual Depth Estimation: Bridging the Gap in 3D Object Detection for Autonomous Driving (CVPR 2019 · 1,193 citations)
- 2020 · HVNet: Hybrid Voxel Network for LiDAR Based 3D Object Detection (CVPR 2020 · 244 citations)
- 2021 · RSN: Range Sparse Net for Efficient, Accurate LiDAR 3D Object Detection (CVPR 2021 · 164 citations)
- 2022 · AFDetV2: Rethinking the Necessity of the Second Stage for Object Detection from Point Clouds (AAAI 2022 · 137 citations)
- 2023 · Omni3D: A Large Benchmark and Model for 3D Object Detection in the Wild (CVPR 2023 · 106 citations)
- 2024 · SAFDNet: A Simple and Effective Network for Fully Sparse 3D Object Detection (CVPR 2024 · 78 citations)
- 2025 · ViKIENet: Towards Efficient 3D Object Detection with Virtual Key Instance Enhanced Network (CVPR 2025 · 10 citations)
- 2026 · SemLT3D: Semantic-Guided Expert Distillation for Camera-only Long-Tailed 3D Object Detection (CVPR 2026 · 0 citations)
Earliest papers here
- Pseudo-LiDAR From Visual Depth Estimation: Bridging the Gap in 3D Object Detection for Autonomous Driving (CVPR 2019 · 1,193 citations)
- HVNet: Hybrid Voxel Network for LiDAR Based 3D Object Detection (CVPR 2020 · 244 citations)
- DSGN: Deep Stereo Geometry Network for 3D Object Detection (CVPR 2020 · 224 citations)
Benchmark records
The best number each year that beat every earlier one, among papers of this approach reporting their own method. Numbers are extracted by AI from each paper's text; check the paper before citing.
NuScenes validation · mAP (higher is better)
4 papers, 2020–2024 · not reported since 2024
| Year | Best so far | Set by |
|---|---|---|
| 2020 | 35.4 | What You See is What You Get: Exploiting Visibility for 3D Object Detection (CVPR 2020 · 135 citations) |
| 2022 | 52.89 | Joint 3D Object Detection and Tracking Using Spatio-Temporal Representation of Camera Image and LiDAR Point Clouds (AAAI 2022 · 20 citations) |
| 2024 | 66.3 | SAFDNet: A Simple and Effective Network for Fully Sparse 3D Object Detection (CVPR 2024 · 78 citations) |
nuScenes test · mAP (higher is better)
4 papers, 2020–2024 · not reported since 2024
| Year | Best so far | Set by |
|---|---|---|
| 2020 | 35 | What You See is What You Get: Exploiting Visibility for 3D Object Detection (CVPR 2020 · 135 citations) |
| 2022 | 62.9 | Boosting 3D Object Detection by Simulating Multimodality on Point Clouds (CVPR 2022 · 35 citations) |
| 2024 | 68.3 | SAFDNet: A Simple and Effective Network for Fully Sparse 3D Object Detection (CVPR 2024 · 78 citations) |
nuScenes val · mAP (higher is better)
4 papers, 2023–2025 · still improving
| Year | Best so far | Set by |
|---|---|---|
| 2023 | 63.6 | LinK: Linear Kernel for LiDAR-Based 3D Perception (CVPR 2023 · 56 citations) |
| 2025 | 68.72 | ViKIENet: Towards Efficient 3D Object Detection with Virtual Key Instance Enhanced Network (CVPR 2025 · 10 citations) |
What has been tried
48 papers here report 541 measured changes (a component added, removed or swapped, compared with a baseline): helped 79%, no clear effect 4%, hurt 13%, mixed 4%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| Replace center-voxel heatmap targets with nearest non-empty voxel targets | Waymo Open · mAPH | 55.2 -> 71.0 | Helped | SAFDNet: A Simple and Effective Network for Fully Sparse 3D Object Detection (CVPR 2024 · 78 citations) |
| Add DMA to PP-IF | DAIR-V2X val · AP@0.5 | +1.05% | Helped | DI-V2X: Learning Domain-Invariant Representation for Vehicle-Infrastructure Collaborative 3D Object Detection (AAAI 2024 · 48 citations) |
| Increase WaffleIron width from 256 to 768 while distilling DINOv2 ViT-L/14 | nuScenes val linear probing · mIoU | 61.6 -> 66.8 | Helped | Three Pillars Improving Vision Foundation Model Distillation for Lidar (CVPR 2024 · 23 citations) |
| add MFC to single-frame clustering | WOD validation, Vehicle · 3D AP L1, IoU0.5 | 17.36 -> 19.91 | Helped | Commonsense Prototype for Outdoor Unsupervised 3D Object Detection (CVPR 2024 · 22 citations) |
| Use DRET-Aug instead of no rain augmentation | WOD-DA rainy, All(L2) · mAP | 55.8 -> 56.6 | Helped | Sunshine to Rainstorm: Cross-Weather Knowledge Distillation for Robust 3D Object Detection (AAAI 2024 · 17 citations) |
| Use dynamic rather than shared f(θ) weights | KITTI val, distant (40m-Inf) · LDS | 32.5 -> 36.2 | Helped | Improving Distant 3D Object Detection Using 2D Box Supervision (CVPR 2024 · 11 citations) |
| Add BEV fusion | KITTI val, car · AP (R40), hard | 83.20 -> 84.71 | Helped | ViKIENet: Towards Efficient 3D Object Detection with Virtual Key Instance Enhanced Network (CVPR 2025 · 10 citations) |
| add Collision Probability Encoding | OPV2V · AP@0.3 | 30.02 -> 55.62 | Helped | Learning to Detect Objects from Multi-Agent LiDAR Scans without Manual Labels (CVPR 2025 · 8 citations) |
| change anchor size at test time | Waymo → KITTI, Waymo → NuScenes, and NuScenes → KITTI · 3D AP | 5.38 / 35.55; 1.31 / 14.29; 15.0 / 27.11 | Helped | An Empirical Study of the Generalization Ability of Lidar 3D Object Detectors to Unseen Domains (CVPR 2024 · 7 citations) |
| Add position features to velocity-only node features | Waymo val pseudo, SegIoU@0.4 · F1 | 67.1 -> 70.9 | Helped | SeMoLi: What Moves Together Belongs Together (CVPR 2024 · 7 citations) |
| Add silhouette loss to projection-loss auto-labeling | KITTI-360 training, Easy AP3D@0.5 · AP3D@0.5 | 23.33 -> 26.84 | Helped | VSRD: Instance-Aware Volumetric Silhouette Rendering for Weakly Supervised 3D Object Detection (CVPR 2024 · 5 citations) |
| Increase backbone and image size for DETR3D (ResNet-50, 704×256 to ResNet-101, 800×450) | UAV3D validation 3D detection · mAP | gains of approximately 0.18 in mAP | Helped | UAV3D: A Large-scale 3D Perception Benchmark for Unmanned Aerial Vehicles (NeurIPS 2024 · 5 citations) |
What reviewers will expect
Datasets and baselines named in the 7 papers of this approach from 2025–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Datasets
- Waymo Open Dataset · 2 of 7 papers
- nuScenes · 2 of 7 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 37 papers; from OpenAlex and the papers' affiliations).
- Xiamen University · 3 papers
- NVIDIA · 3 papers
- Huazhong University of Science and Technology · 2 papers
- Beijing Institute of Technology · 2 papers
- Texas A&M University · 2 papers
- Carnegie Mellon University · 2 papers
- The Chinese University of Hong Kong · 2 papers
- Caltech · 2 papers
Use it
Download BibTeX of 38 key papersOpen the topic in the interactive map
Get alerts. Describe your idea in scime atlas and save it (free scime account): you get a weekly email when new preprints land close to it.
Related approaches
Same keyword cluster
- Unified LiDAR 3D detection (107 papers)
- 3D objectness and scene segmentation (57 papers)
- Monocular 3D object detection (56 papers)
Elsewhere in the topic
- LiDAR perception for driving (75 papers)
- Multi-camera BEV detection (69 papers)
- Camera-radar fusion detection (36 papers)
Benchmark numbers, ablations and baseline numbers are extracted by AI from each paper's full text and can be wrong; the paper is the authority. Data: results index of 2026-11-02, citation counts from OpenAlex (2026-10-09).